2.6 Level 1b Products and Algorithms

2.6.1 Level 1b Algorithms

Most of the contents of this section has been extracted from the “MERIS Level 1 Detailed Processing Model” document ( R-1).

The MERIS Level 1b processing is in charge of reading the MERIS Level 0 product; checking the packets; extracting measurement data from the packets; correcting, calibrating and geolocating the Earth imaging data into spectral radiance values at the top of the atmosphere; ingesting ancillary data; creating Level 1 products which include radiances, geolocation and other annotations. On-line quality checks are performed at each processing stage.

The logic of the Level 1b processing algorithm follows the functional breakdown diagram shown in figure 2.19 below. The same logic applies to RR and to FR processing.

Figure 2.6 - Functional breakdown.

2.6.1.1 Source Data Packet Extraction

MERIS Level 0 processing is assumed to sort packets in the data stream which correspond to the observation modes of MERIS, from those corresponding to onboard characterisation modes.

At the initial stage of Level 1b processing, information in the packet header and data field header is used to detect in the FR or RR stream of packets such anomalies as:

· transmission error

· format error

· sequence error

The onboard time code needs to be converted to Universal Time (UT) for datation of the packets acquisition.

2.6.1.2 Saturated Pixels

MERIS samples may be affected by phenomena outside the range of the useful measurements; i.e., a spectral radiance between 0 and L_sat. Such samples are totally invalid, the corresponding cells being affected temporarily or permanently. When possible, invalid pixels should be replaced by a good estimate.

Such phenomena are:

1. Saturation by radiance level above L_sat (caused by, e.g., Sun glint, cloud, bright land or snow /ice), which affects cells temporarily (typically several columns in several bands over several frames).

2. Recovery from saturation. After saturation, components of the acquisition chain need some time (a few pixel columns) to recover; in the meantime the measurement is affected.

3. Blooming. Samples in bands and columns close to a saturated one may be temporarily affected by photon or photoelectrons diffusing from the saturated pixel.

4. Glitches, high-intensity impacts (e.g., laser) generate isolated high value samples.

5. Dead pixel. Due to manufacturing defects or to ageing in space, the response of some CCD cells to light may "die;»I.e., permanently deviate too much (to the extent that gain correction is not usable) from the useful measurement range. Such dead pixels need to be known.

Samples affected by saturation/recovery/blooming (1, 2, 3) are flagged.

Samples corresponding to dead pixels (5) are replaced with a cosmetically interpolated value after radiometric calibration within the radiometric processing step.

Glitches are neither detected nor corrected due to unavailability of a simple model for detection.

2.6.1.3 Radiometric Processing

The valid MERIS samples are digital counts resulting from the acquisition by MERIS of passive optical spectral radiance remote sensing. The objective of the radiometric processing is to estimate the spectral radiance which caused these counts. An inverse model of the MERIS acquisition is used for that purpose, using parameters stored in the Characterisation data base and the MERIS samples themselves. The MERIS acquisition model is described as:

where:

X_b,k,m,f is the MERIS raw sample (not corrected onboard)

NonLin_b,m is a non-linear function

T_f ^VEU is the amplification unit temperature

T_f ^CCD is the sensor temperature

g(T) and g_c(T) are temperature-dependent gain terms (close to 1)

A_b,k,m is the "absolute radiometric gain"

L_b,k,m,f is the spectral radiance distribution in front of MERIS

S_b,k,m,f is the smear signal, due to continuous sensing of light by MERIS

G_b,k,m is a linear process representing the stray light contribution to the signal. For a given sample, some stray light is expected from:

· all the other simultaneous samples in the module, spread into the sample by specular (ghost image) or scattering processes

· the samples in the previous and following frames

C⁰ _b,k,m is the dark signal (corrected onboard for temperature effects by the Offset Control Loop)

is a random process representative of the instrument errors and parasitic processes not accounted for in the other terms of the model.

All terms not indexed by f (frame) do evolve in time due to ageing, but with a much slower rate which allows to represent them, for a given Level 1b product, as fixed quantities retrieved from data bases.

The radiance sensed by MERIS L_b,k,m,f is, for a given set of target physical parameters and illumination and observation angles, proportional to the extraterrestrial Sun spectral flux. Because there is no absolute spectral measurement of the Sun irradiance simultaneous to MERIS acquisition, all results are produced with reference to a Sun spectral flux model which must be included in the product header.

The term A_b,k,m reflects all the amplification gains inside the instrument, which depend on:

· instrument programming (band settings, amplification programmable gains)

· components ageing

· components temperature

· power supply voltage

In order to provide for limitation or failure of the onboard temperature regulation, there shall be a residual correction for g(T), GC(T). In normal operation, T depends on the time elapsed since the Sun zenith angle has decreased below a threshold (80°) and can be predicted.

2.6.1.4 Stray Light Correction

The stray light term G_b,k,m(L_*,*,*,*) in the MERIS acquisition model above may be strong enough to affect the Least Significant Bits of the raw data. This may happen in particular when MERIS is observing a scene with some high radiance areas (Sun glint patch, partly cloudy ...). As the linear transform G_b,k,m is assumed to be known well enough from instrument characterisation, it is possible to compute an estimate of the stray light, and correct for it.

Stray light correction is handled separately from radiometric processing due to the specific nature of the processing in that stage: de-convolution; and to the fact that it can be switched on and off.

2.6.1.5 Geolocation

The geolocation problem encompasses all processing that is directly related to the location on Earth of the MERIS measurement data.

The points where the MERIS radiance samples have been measured are determined by the projection on Earth of the line of sight of every pixel. That projection depends on

· the shape of the Earth

· the altitude of the sample

· the position of the ENVISAT satellite at the time of acquisition

· the orientation of the MERIS modules

· the optics of each MERIS module

In order to simplify product handling, the MERIS radiance samples are relocated by nearest neighbour interpolation to the MERIS product grid, which has the following characteristics (FR grid):

· central column: subsatellite point track on Earth

· line orientation: perpendicular to spacecraft velocity, projected on Earth

· columns spacing: fixed for one product, 260 m (with very small variations)

· number of columns: 4,481

· line spacing: variable with time and orbit altitude, fixed by the MERIS frame time of 0.044s (mean » 292 m)

The RR-grid is a 4 x 4 subsampled version of that grid.

The surface of altitude 0 on Earth is approximated by a geoid model. The model WGS-84 used by the ENVISAT orbit propagator shall be used.

Knowledge of the ENVISAT platform and attitude relies on:

· prediction or estimation of the satellite position and attitude; the ESA CFI software is used:

- po_ppforb or po_interpol for orbit propagation

- pp_target for attitude modelling

· accurate datation of the MERIS samples, to the MJD2000 time reference used by the orbit and attitude prediction/estimation.

The interpolation algorithm for resampling MERIS data to the grid may use characterisation data defining the MERIS pixels de-pointing. Neglecting the surface elevation causes an error in pixel location, proportional to altitude and to the tangent of the observer zenith angle. That error is estimated at the tie points.

Sun zenith and azimuth angle, observer zenith and azimuth angle, may be computed for any pixel knowing pixel location and Sun direction in a common frame but are stored only at the product tie points.

Sun glint, because of the high radiance values measured there, has an impact on both the direct usage of L1b data and on L2 processing. A first estimate of the affected pixels is performed. The location of the potential Sun glint can be predicted for each pixel, from the illumination and observation geometry.

Geolocation processing is broken down into 5 main algorithm steps:

· product limits

· tie points Earth location

· altitude retrieval

· resampling

· Sun glint

2.6.1.6 Pixel Classification

In order to make easier the exploitation of TOA radiances by further processing (e.g., Level 2, Browse), the Level 1 product contains appended information about the nature of each MERIS pixel. The classification process uses the a priori knowledge of a land/ocean map indexed by longitude and latitude, and the information in the TOA radiance bands to classify each valid pixel into:

· clear sky/ocean

· clear sky/land

· bright pixel/ocean

· bright pixel/land

Bright pixels include clouds, bright sand or soil, ice, snow, Sun glint...; the a priori known nature of the underlying surface is kept.

Clear sky is to be understood as clear enough to pursue atmospheric corrections.

2.6.1.7 External Data Assimilation

In order to make easier the exploitation of TOA radiances by further processing (e.g., Level 2), the Level 1 product contains appended information about the environmental conditions prevailing at the time and place of the MERIS acquisition. The parameters of interest are:

· atmospheric pressure at surface level for prediction of the Rayleigh reflectance, optical thickness

· surface wind speed and direction for prediction of Sun glint and whitecaps

· relative humidity at 850 hPa for verification of the aerosol correction

· total ozone column contents for atmosphere absorption correction

These parameters are acquired from external source (ECMWF data) and are interpolated, space-wise, to the tie points.

2.6.1.8 Formatting

All the data and flags derived in the above algorithm steps are formatted into a file compliant with the Level 1b product description found in the Formats section of this handbook.